• Retrieval Augmented Generation (RAG) is a technique popularized by Meta in 2020 that boosts the performance of Language Models by providing relevant context to the model alongside the question/task details.
  • The operation of RAG involves three stages: data preparation, retrieval, and generation. It can be used to improve Language Models' performance on tasks like summarization, translation, etc., and is particularly useful when the dataset is dynamic or not large enough.